Thank you for pointing to the Giulio Antonini's paper on the
Sqrt(f)*(1+j)*Rs skin-effect equation in July issue of IEEE Transactions
on MTT. According to this paper, the error in the inductance introduced
by this equation is about 30% at higher frequencies for rectangular
geometries. This obviously translates into delay error. But even
larger error is created by its singular dc inductance.
The Sqrt(f)*Rs skin-resistance equation with frequency-response
correction that I used in W element before it was changed to
Sqrt(f)*(1+j)*Rs in 99.2 Hspice release is valid for any geometry at
high frequencies. And the inductive component that I restored by
frequency-response correction was well-behaved at low frequencies.